To detect computer code problems, Intel has just released its machine learning-based verification and testing tool in open source. The first feedback shows good performance.
With the growth of applications, analyzing code has become a challenge for developers to ensure the quality of code. Intel has created a machine learning-based tool to address this issue. The solution is called ControlFlag and has just been opened to the community by the founder of Santa Clara.
ControlFlag works with several programming languages that contain control structures (that is, blocks of code that specify the flow of control in an application). The tool aims to reduce debugging work by taking advantage of unsupervised learning. This is an algorithm subject to “unknown” data for which there are no previously defined categories or labels. ControlFlag’s machine learning system, in this case must learn to classify data, processing unlabeled data to learn from its inherent structure.
The first tests of ControlFlag are encouraging. “Last year, he identified a code anomaly in Client URL (cURL), a computer software project that transfers data using various network protocols over a billion times a day,” says Justin Gottschlich. , director of Intel Lab. “Likewise, ControlFlag found dozens of unprecedented anomalies in several open source software repositories,” adds the manager.
Demand for the quality of code is growing rapidly and is attracting several developers. A 2020 Gartner study found that IT companies spent around $ 2 trillion on software development, and 50% of budgets are related to code debugging. This market is developing more and more and the competition promises to be tough with several players such as Tabnine, Ponicode, Snyk and DeepCode. We can also mention IBM’s OpenAI which also relies on machine learning in application development.